Your Resource for All Things Apps, Ops, and Infrastructure

Cisco HyperFlex: 1 Year Later

As the Converged (CI) and Hyperconverged (HCI) infrastructure practice lead at AHEAD, I spend a significant portion of my time evaluating and designing solutions, including Cisco’s HyperFlex (HX). I wrote this introductory post shortly after HX was released in 2016, but before I received an HX unit in our lab. At that time, almost everyone’s HyperFlex product knowledge was based on PowerPoints, controlled remote lab experiences, or customer feedback.

Fast forward roughly 1.5 years later (and a few days after Cisco Live 2017), where I’ve had the opportunity to talk to dozens of customers about HX, run HX through the paces in AHEAD’s lab, work with customers on POCs of HX in their IT environments, and design HX solutions for production workloads.

I reached out to Chris Nichols, Sr. Systems Engineering Manager, responsible for global field engineering for HyperFlex, to get his opinion on why HX continues to be top of mind for organizations that are exploring HCI.

“HyperFlex has been a tremendous success by building on an already successful and proven UCS compute platform, the key to all of today’s Converged Infrastructure designs – many of which AHEAD and Cisco have been building hand-in-hand.”

– Chris Nichols, Sr. SEM, Global Systems Engineering, Cisco

This post focuses on the evolution of HX since its original launch in early 2016. I discuss both positive and negative elements of the technology, appropriate use cases, and HX’s relevance in the HCI market.

How it Began (HX 1.7)

When HX was first released to the public in February of 2016, many customers, and all HCI competitors, pegged it as a non-scalable, feature-poor, inflexible HCI solution. Based solely on what the product’s capabilities at the time, I agreed with them. That said, I also saw the promise for a much better product, and understood that Cisco’s conservative nature was limiting the capabilities of HX to ensure a positive customer experience, as HX proved in the field.

Some of the specific attributes of the product that raised alarms for many included:

  •  8 node cluster maximum
  •  vSphere only support
  •  No encryption, replication, or backups
  •  Sluggish web plugin
  •  Lack of single-click upgrade (at least a fully encompassing button)
  •  Heavy storage controller overhead
  •  Lack of hardware flexibility (only support for 1.2TB drives, all nodes have to be identical)
  •  No All Flash options
  •  No Gfx Option, no nested vCenter support

Requires Fabric Interconnects, making it less attractive for Robo

Today, these attributes look pretty grim when you put current generation HCI products like VxRail or Nutanix next to them.

Now, let’s look at where HX is right now, and where it will be by the end of the year.

HX 2.5

  • SED encryption support, with external key management
  • Native snapshot-based replication w/ 15M RPO
  • New HTML5 management GUI
  • Proven single click upgrade (still requires interaction with FIs)
  • Storage controller overhead has not increased since product launch
  • Hardware Flexibility – Now supports 3 drive sizes including all flash
  • Hardware Flexibility – Supports M60 and M10 cards
  • Hardware Flexibility – Expanded support of B200 M3/M4, B260s, B420/460 M4s, C220 M3/M4s, and C460M4s

HX Features slated for 2017

  • Microsoft Hyper-V Support
  • NVME Caching, M5 Skylake-SP models
  • 64 Node clusters (32+32)
  • Many to 1 native replication support
  • Erasure Coding
  • Fault Domain Support
  • Starship – Cloud-based HTML5 management of multiple HX Clusters

Looking at HX now, and in the near future, you can see that Cisco has not fallen behind. In fact, when it comes to available features, Cisco has shortened the distance between HX and its primary competitors.

The speed at which Cisco has been able to release updates to the HX platform is astounding, but it is par for the course for the HCI leaders in the market. If you were an early adopter in HX, the system that you purchased 18 months ago now has a considerable number of additional features, with more coming over the next few months. This is all for free if you maintained software support.Try doing that on a storage array…

Use Cases

HX was originally promoted by Cisco to focus on a limited number of use cases — basically light virtualized workloads that were not too demanding. This was not due to the lack of performance or scalability in the architecture, but instead to Cisco’s conservative nature on a relatively new domain, HCI.

Today, however, HX is recommended for just about any workload who’s operating system it supports.

ROBO is attainable now that you can run HX with 3 nodes and without a pair of FIs. It will have a much more compelling story once Project Starship launches the cloud based HTML web interface to manage all your remote locations (think of it like Meraki for HX)

VDI has always been attainable, and it is simply even better now with all flash support, and M10 and M60 support. A unique feature of HX for VDI is that for storage-light workloads, HX shines by allowing compute-only nodes, enabling customers to potentially build a more efficient HCI cluster than it’s competitors through a combination of mixed and compute-only nodes

From the launch of HyperFlex, we were very bullish on the value of including networking as part of the HyperFlex design, the performance advantages of our software defined storage layer, and the flexibility we gave our customers, mixing both HX and compute nodes in a cluster.”

– Chris Nichols, Sr. SEM, Global Systems Engineering, Cisco

Databases are a new addition to HX’s workload stable since the all flash release on 2.0.

General Virtualization is and has been doable since the product’s launch and today with much larger cluster sizes and native support for SEDs and upcoming support for native replication makes this use case even easier.

This wouldn’t be anything more than a regurgitation of product features if I didn’t take the time to give my opinion on how HX compares to the competition.



HX is the most performing HCI platform in the market when compared to others using similar platforms (except for, perhaps, VxRack FLEX) based on tests that we have done and results we have seen at our customer’s locations. This is because of HX’s parallel architecture.

Technically, we have seen Nutanix win on read latency, but as a whole — so far — HX wins the performance battle, especially prior to spending time tuning the system and the VMs.


Scalability used to be an issue for HX because you could only have 4 HX clusters under management in vCenter, and each cluster could support at most 8 HX nodes. Now, you can have 32 HX nodes along with 32 compute nodes. And there you have it, vSphere’s cluster maximum supports 64 nodes!

“The flexibility HyperFlex allows by scaling storage and compute independently, and even integrating their existing storage arrays, also allows them to evaluate how their on-demand infrastructure needs can evolve and grow over time, directly aligning with their business needs.”

– Chris Nichols, Sr. SEM, Global Systems Engineering, Cisco

Manufacturer Ecosystem 

If you value UCS, you will probably value HX with all the flexibility, visibility, and manageability that UCS Fabric Interconnect brings.

Furthermore, we expect to see Cisco continue to integrate HX into their other product lines that include UCS Director and Cloud Center

Address the Network 

If HCI scares you because you think it puts too much pressure on an optimal network, then HX might be for you. HX requires the use of fabric interconnects which address network aggregation for the HX nodes, and the automated install wizard will provision all the QoS policies for you.


Atypical for Cisco, HX appears to be very aggressively priced.

Cons / Lacking Features

Replication / Improved Replication

It is on the roadmap, but initial native replication will have limited configuration options with a 15M RPO.

Integration with Cloud

No ability to replicate snaps or VMs to one or more cloud providers

Multi-Hypervisor Support

Hyper-V is coming, hopefully this year. KVM would also be interesting to see.

Simplified Management

UCS Manager is great at giving you ultimate control and configuration flexibility, but that also makes it complicated if you aren’t familiar with UCS. This is a hard issue to solve at a time where many HCI competitors are focusing on simplicity, and today, HX has two management screens, the new HX HTML5 GUI and UCS Manager.

Erasure Coding

This is also on the roadmap and very important for customers with data that doesn’t cooperate with Deduplication and Compression, such as databases, encrypted data, and/or pre-compressed data.

Putting it all together

After all is said and done, Cisco HyperFlex is a viable competitor in the HCI market, which is one of the few IT infrastructure segments that is seeing large growth. Recent data put Q1 2017 worldwide HCI revenues at about $665M, about a 65% increase against Q1 of 2016. At the same time, worldwide server sales, especially blade servers, are shrinking, so Cisco is very motivated to not only stay relevant, but to win in the HCI category.

What we’re most excited about, though…is how fast Cisco has been able to step into this market, execute on our planned roadmap, deliver product value, and make HyperFlex a cornerstone of our Data Center portfolio.

– Chris Nichols, Sr. SEM, Global Systems Engineering, Cisco

So far HyperFlex has roughly 1,600 customers, which is less than VMware VSAN or Nutanix. They do, however, have 60,000~ UCS customers, many of which probably invested in UCS because they liked fabric interconnects and UCS Manager.

“…HyperFlex has also brought hundreds of net-new UCS customers to the Cisco family. That tells me it’s not just a great message on a slide; it’s real results our customers see, evaluate, and ultimately decide to purchase and use for their environments.”

– Chris Nichols, Sr. SEM, Global Systems Engineering, Cisco

With these 60K UCS  customers comes a tremendous advantage since they are already using the core infrastructure component of HX. Now Cisco just needs to convince them to move from UCS only to UCS + HX in their environment.

The challenge, which is a similar challenge to the converged infrastructure and reference infrastructure manufacturers, is that it is difficult to promote the simplicity of a product when it has hundreds, even thousands, of toggles and buttons.

As successful as both UCS and HX have been, they are both perceived to be complicated products, and since HCI is predominantly associated with simplicity, Cisco needs to continue to focus on simplifying HX’s operations if they want to win. This is a tricky—but not impossible—challenge, because they need to walk a tightrope of making HX look less like UCS, while at the same time not alienating their UCS customers. A path forward could be to continue automating the HX experience without dumbing down UCS.

We will see if project Starship takes them there…

To learn more about CI and HCI products and strategies, contact us today to meet with our experts one-on-one and for more information on our current offerings and solutions.

Subscribe to the AHEAD i/o Newsletter